123755 - bad behavior with opacity, sidebars, and window resizing

Status: RESOLVED FIXED

(SeaMonkey :: Sidebar, defect)

Description

[Bill Nottingham]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.8) Gecko/20020205
BuildID:    2002020511

I'm seeing weird behavior with the combination of:

1) opacity (set via a style sheet)
2) sidebars that contain HTML (not XUL) content
3) resizing of the mozilla window

Using Linux, 0.9.8. Mozilla either crashes, or overwrites the
sidebar display.

Reproducible: Always
Steps to Reproduce:
1. Have a chrome/userContent.css that contains:
BODY { 
     -moz-opacity: .50 !important;
}

2. Have a default sidebar panel that contains HTML (for example,
   point it at http://bugzilla.mozilla.org/)
3. Open a default-sized window, and go to a web site (for example,
   http://www.redhat.com/)
4. Hit the 'maximize' button in your window manager of choice

Actual Results:  It varies depending on the sidebar content, AFAICT. Sometimes
the contents of the sidebar are overwritten with stuff from the web
page. I've also seen mozilla crash. If the sidebar is overwritten,
refreshing it works.

Expected Results:  Not crashed, and not overwritten the sidebar display.

I'll try and get a backtrace of the crash, once I get a debugging build going.

Comment 1

[Bill Nottingham]

CC:ing blizzard, since the crash is in the GTK code, in case he has any ideas
what's going on. ;)

Comment 2

[Bill Nottingham]

Attachment: gdb traceback of mozilla crash

This is one of the cases where it crashes. The sidebar content in this case is
a random HTML file I wrote (basically, bookmarks + some form stuff)

Comment 3

[Christopher Blizzard (:blizzard)]

Not my usual area, adding the image posse.

Comment 4

[tor]

Looks like nsBlender - adding roc.

Comment 5

[Samir Gehani]

Over to pavlov for a look.

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

I bet it's me :-).

BTW adding
> BODY { 
>     -moz-opacity: .50 !important;
> }
to your userContent.css is crazy :-)

Comment 7

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Partial fix

The problem seems to be that we're running out of resources and various things
are failing silently. I haven't been able to pin down exactly which things.
Running out of resources is not surprising; when I maximise my window with
these opacity settings, the view manager tries to allocate 3 2048x1024 pixel
offscreen drawing surfaces.

This patch prevents the crash, at least. nsDrawingSurfaceGTK::Lock must check
to see if it got the image; if not, then we just return failure. The view
manager detects this and paints opaquely instead. The "Blend failed!" warning
IS triggered sometimes.

The patch also cleans up some error handling in
nsViewManager::CreateBlendingBuffers(). However, I still sometimes see the
sidebar (or the content area) not being painted properly sometimes. I presume
some GTK calls are failing silently, but I'm not sure which ones.

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

blizzard, pavlov, can you sr/r this respectively?

Comment 9

[Stuart Parmenter]

Comment on attachment 69003 [details] [diff] [review]
Partial fix

r=pavlov

Comment 10

[Stuart Parmenter]

-> roc

Comment 11

[Marc Attinasi]

Comment on attachment 69003 [details] [diff] [review]
Partial fix

sr=attinasi

Comment 12

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Checked in. I'm leaving this bug open because things still aren't rendered
properly in some cases and I don't know where the failing GTK call is.

Comment 13

[Brant Gurganus]

By the definitions on <http://bugzilla.mozilla.org/bug_status.html#severity> and
<http://bugzilla.mozilla.org/enter_bug.cgi?format=guided>, crashing and dataloss
bugs are of critical or possibly higher severity.  Only changing open bugs to
minimize unnecessary spam.  Keywords to trigger this would be crash, topcrash,
topcrash+, zt4newcrash, dataloss.

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Marking FIXED because the original bug was fixed and all the relevant code has
changed a lot since this bug was reported. If there are any remaining issues
someone else will report them :-).